There are active discussions around both meta and several TCS questions about what questions should be rejected. The two reasons most often stated are (afaik)

  • too basic (aka too localized, whatever that means)
  • homework problem

There is a multitude of opionions about wether or not these reasons are legitimate or not; I do not want to discuss this here. Let us assume they are (seems reasonable since there is a large faction in favor of them). But then, should we not also close questions like these:

My rationale: These and other questions that are phrased like "What is best ... known?" or "Is there ...?" are "only" a replacement for literature research. They do not pose a motivated, original problem or reflect that work has gone into finding an answer. One could say the one asking has not done his homework.

Of course they are research level questions and therefore inherently more interesting -- and they offer a possibility for experts to show off (in a good way), other than undergrad homework problems.

Nevertheless, I feel that a general rule of thumb for question scope should be found and that the negative approach (finding criteria for closing/rejecting) cannot go anywhere.

So, let us try to answer the following questions: What makes a question good? What questions should not be closed?

  • 1
    $\begingroup$ We already have meta.cstheory.stackexchange.com/questions/300/… $\endgroup$ Commented Oct 25, 2010 at 19:36
  • 3
    $\begingroup$ "Localized" is the wrong reason to close a "too basic" question. Questions that are too simple are "off topic" per the scope of the site in the FAQ. $\endgroup$ Commented Oct 25, 2010 at 20:07
  • $\begingroup$ Jukka, I had hoped to influence the "when to close", not the "what to write" discussion, even though I kind of propose the latter as an answer for the first. $\endgroup$
    – Raphael
    Commented Oct 25, 2010 at 21:51
  • $\begingroup$ @Aaron: Read why we should use “Too localized” to close too basic questions. If you still have an objection, I am happy to hear it on that page. $\endgroup$ Commented Oct 26, 2010 at 0:34
  • $\begingroup$ Jukka’s comment answers at least the part “What makes a question good?” $\endgroup$ Commented Oct 26, 2010 at 0:40
  • 2
    $\begingroup$ @Tsuyoshi: I have read your question. I upvoted it, because I thought it was a good thing to discuss, but I also upvoted the comments that said we should use "off topic." Ultimately, I agree with Suresh's comment there, that anyone voting to close, or downvoting a question, should explain why in a comment, or at least upvote someone else's comment if that person got to the reason first. $\endgroup$ Commented Oct 26, 2010 at 0:41
  • $\begingroup$ @Aaron: As I said, I am happy to hear it on that page. This is not the right place to discuss that. Let’s not pollute a discussion with a tangential issue. $\endgroup$ Commented Oct 26, 2010 at 0:42

3 Answers 3


In MathOverflow, some of the most useful answers are pointers to a reference. The math literature is so huge that no one researcher can possibly know more than a small fraction of it, and even with Google, navigating the literature in an area that isn't your own is really, really difficult. Somebody who knows the area can give you an appropriate reference in a few minutes, when it could easily take you hours or days to find it on your own.

The same thing is true for Stack Overflow. Programming also has a huge knowledge base, and nobody knows more than a fraction of it.

Is this not also, to a more limited extent, the case for theoretical computer science? If not, why do we need a StackExchange website?

On the other hand, if the answer is given in the Wikipedia page for the obvious subject to check, the OP is not really trying hard, and we should close the question.

  • $\begingroup$ I strongly agree, as do, I think, most everyone who participated in the private beta and tried to hash out the scope of the site. There is, in fact, an explicit tag ("reference-request") for this type of question. Also, when I have made a reference request, I've put it in the title of the question, e.g., "Reference Request: Characteristics of Fubars." $\endgroup$ Commented Oct 25, 2010 at 23:19
  • $\begingroup$ For some questions, it's not immediately obvious whether there's a reference or not. $\endgroup$ Commented Oct 25, 2010 at 23:21
  • 4
    $\begingroup$ Sure. You recently asked two such questions yourself, I believe. :-) My point was that even a question like, "I know there's a paper about this because I read it five years ago but I can't remember the title or the authors, can anyone help me?" is 100% inbounds, as I understand the scope of the site. $\endgroup$ Commented Oct 25, 2010 at 23:26
  • $\begingroup$ My impression is that this is exactly what StackExchange was designed for. $\endgroup$ Commented Oct 27, 2010 at 20:22

My 2 cents :

  • No question should be deleted except if it is total nonsense (or if it contains questionable content - I mean something that can bring "legal" problems)
  • Making strict rules is not good at that point of the site's life. If it is too strict, people won't ask question. We have a voting mechanism, almost nobody votes currently, but it has been designed so that the community can sort out what is a legit question and what is not, without becoming an oligarchy. So, maybe we can try to vote more often and see what happen.

I have another comment on this point:

The question should not be answerable by only two references

This is a sensitive point, very often I rush into the coffee room and ask to the audience (mainly PhD students of course, they are the only ones always in the coffee room, and also those who know the best the latest papers -> joke) about possible references on the topic that interests me at that time. I think cstheory has also this purpose. And even a top notch google user cannot beat a crowd of TCS researcher trying to understand what you are approximately saying.


For a question to qualify as worthy, I want to read the following (let P be the asking person):

  • Motivation: Why is this question important for P? Note that curiosity or assignment are valid reasons, imho.
  • Clear Problem: The problem/question should be stated in a clear way, both language- and layout-wise. Non-standard terminology (read: not taught at university) should be explained or referenced. Event though experts who are most likely to answer appropriately might know what P is talking about, why not make the question accessible to other people, too?
  • Originality: The question should not be answerable by only two references. Some thought should be required before answering. Note that I consider asking for an explanation might qualify since explaining/representing is often a creative act in itself.
  • Own Work: Depending on the kind of question this can be a commented list of references, a failed/partial proof or general thoughts.

In short: I want to see that P invested time on himself and that there is something in his post that people can refer to, e.g. "You approach can not work because..." or "Reference X does not apply since...".

I realize this means work but this is only fair. It should not take much less time for P to state his question than for any expert to answer properly.

P.S. It is a pity that stackexchange discourages discussion by design. Many questions should really be (and some are) tackled iteratively.

  • 6
    $\begingroup$ I think your guideline is far too strict. Questions of the form "Here is a problem formulation, does anyone know if it has been studied in literature" are perfectly on-topic here. Without knowing the name of the problem it is extremely difficult to google it. Nevertheless, an expert can answer the question in 30 seconds by giving one reference, and I don't think there is anything wrong with that - on the contrary, getting answers to that kind of questions quickly is the strength of this site! $\endgroup$ Commented Oct 25, 2010 at 19:43
  • $\begingroup$ Agreed, valid scenario. Can you provide an own answer with a complete guideline you think appropriate? $\endgroup$
    – Raphael
    Commented Oct 25, 2010 at 21:50
  • 4
    $\begingroup$ I agree with Jukka, and I think Raphael is trying to fix something that is not broken. $\endgroup$ Commented Oct 25, 2010 at 23:21
  • 3
    $\begingroup$ While perhaps a little strict, Raphael's answer doesn't warrent so many downvotes. Rather than required elements of a question, I like to read this as a list of desired properties. For example: some questions are interesting in their own right, or have obvious motivations while others might need some explicit motivation written. $\endgroup$
    – Ross Snider
    Commented Oct 26, 2010 at 5:14
  • $\begingroup$ @Raphael's list forms a great set of objectives, but as @Ross says, these criteria should not be regarded as necessary. As @Jukka points out, we have had really useful one-line responses, some of which were heavily upvoted. $\endgroup$ Commented Oct 26, 2010 at 11:30
  • $\begingroup$ My target here would be agreeing on a list of minimal aspects a question should have in order to qualify. Obviously, quite a number of people want plain reference-asking questions while I was thinking more of substantial questions. $\endgroup$
    – Raphael
    Commented Oct 26, 2010 at 17:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .